Lift Ignitor

AI-driven Recommendation System for Ecommerce & Media Agencies

schedule A Meeting with our Experts

Overview

Folio3 helped implement a solution and provide consequent infrastructure management services for an AI-Driven Recommendations System, a machine learning-driven platform providing real-time personalized content recommendations to optimize and enhance user experience in the Ecommerce & Media/Entertainment industry. This was achieved through a Cloud-first approach through Google Cloud Platform (GCP).

The Challenge - Real Time Data Visualization

Folio3 recognized the customer’s need for powerful infrastructure to accommodate real time data processing, that could:
  • Store & analyze large-scale datasets in real time, with flexible visualization options
  • Ensure a robust infrastructure that allowed seamless log monitoring
  • Ensure stable, high-frequency deployments on the Cloud with minimal downtime and automated backups

The Solution - Cloud-based, flexible infrastructure

The Folio3 team proposed Google Cloud Platform (GCP) as a Cloud-based solution that would allow usage of powerful advanced data analytics tools to achieve real time data processing and visualization.

Adopting a Cloud-first approach for the AI-Driven Recommendations System also allowed Folio3 to tackle these wider challenges:

  • Usage of BigQuery, a fully managed, scale and serverless Data Warehouse that can analyze massive datasets quickly using SQL queries
  • PubSub was used for large-scale activity logging and cleaning training data
  • Compute Engine for backend service components
  • Bigtable and Aerospike were used as NoSQL Database Services for their capability to handle large-scale data and reliability
  • CloudSQL for fully managed database services, along with automated backups, replication and scaling, in addition to easy integration with other GCP services used
  • Bucket (Cloud Storage) for storing and accessing unstructured data
  • Apache Spark was used to process large-scale data and model training through its high-performance distributed computing framework
  • Prometheus for real-time monitoring and alerting
  • Terraform for a consistent and organized infrastructure setup, with streamlined deployments
  • Zipkin for tracking end to end requests
  • Apache Thrift for communication between different services
  • Hotjar for session recording and investigating reported issues
  • PagerDuty for incident management and response

Tools & Technologies Used

Infrastructure Stack